Multimodal integrated approaches in low grade glioma surgery

Surgical management of Diffuse Low-Grade Gliomas (DLGGs) has radically changed in the last 20 years. Awake surgery (AS) in combination with Direct Electrical Stimulation (DES) and real-time neuropsychological testing (RTNT) permits continuous intraoperative feedback, thus allowing to increase the extent of resection (EOR). The aim of this study was to evaluate the impact of the technological advancements and integration of multidisciplinary techniques on EOR. Two hundred and eighty-eight patients affected by DLGG were enrolled. Cases were stratified according to the surgical protocol that changed over time: 1. DES; 2. DES plus functional MRI/DTI images fused on a NeuroNavigation system; 3. Protocol 2 plus RTNT. Patients belonging to Protocol 1 had a median EOR of 83% (28–100), while those belonging to Protocol 2 and 3 had a median EOR of 88% (34–100) and 98% (50–100) respectively (p = 0.0001). New transient deficits with Protocol 1, 2 and 3 were noted in 38.96%, 34.31% and 31,08% of cases, and permanent deficits in 6.49%, 3.65% and 2.7% respectively. The average follow-up period was 6.8 years. OS was influenced by molecular class (p = 0.028), EOR (p = 0.018) and preoperative tumor growing pattern (p = 0.004). Multimodal surgical approach can provide a safer and wider removal of DLGG with potential subsequent benefits on OS. Further studies are necessary to corroborate our findings.


Materials and methods
Study population. Data from adult patients who underwent surgery for DLGG in a single institution between 2000 and 2018 were retrospectively analyzed.
Patients were enrolled according to the following criteria: • Age ≥ 18 years • Preoperative MRI suggestive of supratentorial LGG near or involving the cortical and subcortical eloquent areas • No previous surgery • No preoperative chemo-or radiotherapy • At least 18 months of follow-up • Objective evaluation of EOR on MRI images in DICOM format based on T2-weighted MRI sequences • Revision of histopathological specimens according to the 2016 WHO Classification of Tumors of the Central Nervous System • Intraoperative brain mapping and neurophysiological monitoring.
Needle biopsies were excluded from the study. The tumor involvement of a cortical or subcortical eloquent was detected according to the relationships of DLGG to eloquent regions of the brain as predicted by the preoperative magnetic resonance imaging scan, functional MRI and Diffusion Tensor Imaging (DTI) data.
The local Ethics Committee, Comitato Etico Unico Regionale del Friuli Venezia Giulia, approved this investigation (protocol N.0036567/P/GEN/EGAS, ID study 2540). Considering that the study was retrospective, written consent to participate in the study was not applicable. Written informed consent was obtained for surgery. Consent in this form was decided and approved by the local ethics committee Regione Friuli Venezia Giulia, Italy. All research was performed in accordance with the Declaration of Helsinki.
Intraoperative surgical protocol. The surgical procedures were conducted under cortical and subcortical white matter brain mapping, according to the intraoperative technique previously described 19 .
The AS protocol was selected following the standard protocol in use during that specific time period at our institution 17 . In addition to DES, RTNT was applied in the last series during AS 17 .
In this investigation, the data were collected over about 20 years (from 2000 to 2018), during which the intraoperative technical protocol changed.
The following 3 consecutive intraoperative protocols were used during this time period: Patients belonging to Protocol 1 and 2 were operated on AS or General Anesthesia (GA) based on the specific clinical and neuropsychological status of each subject. Volumetric analysis. Magnetic resonance images in DICOM format were used to assess both pre-and postoperative tumor volume by using axial preoperative post-contrast T1-weighted MRI slices, pre and postoperative axial T2-weighted MRI studies. All pre-and postoperative tumor segmentations were performed manually across all MRI slices using the OsiriX software tool 20 .
The preoperative tumor growing pattern (infiltrative vs. expansive) was expressed as ΔT2T1 index that is preoperative volumetric tumor volume on T2-weighted MRI images-preoperative volumetric tumor volume on T1-weighted images 6,20 . As previously described, higher levels of preoperative ∆T2T1 MRI index represent the prevalence of the diffusive growing mechanism with the tendency to infiltrate the functional cortical areas and subcortical pathways, thus limiting the achievable resection.
The postoperative EOR was evaluated by using 3D T2-weighted MRI axial images as follows: (pre-operative tumor volume -post-operative tumor volume)/(pre-operative tumor volume) 20 .
Histological and molecular analysis. All histological samples were reviewed according to the 2016 World Health Organization (WHO) classification 21 . Molecular markers were evaluated as previously described 22 . Statistical analysis. Statistically significant differences on distribution were evaluated performing chisquared test for categorical variables, and t-test, Wilcoxon rank-sum tests or Kruskal-Wallis for continuous variables, as appropriate.
Overall Survival (OS), progression free survival (PFS) and malignant progression free survival (MPFS) were estimated using the Kaplan-Meier approach. Univariate and multivariate Cox regression models were performed to identify the association between any variable and OS, PFS and MPFS as outcome variables, after the proportional hazards assumption had been verified.
The primary endpoints were EOR and OS differences among the three groups.
Retention in the stepwise model required the variable to be significant to a level of p = 0.05 in a multivariate analysis 23 .
The Spearman's rank correlation coefficient was computed to define the relationship between pre-operative ΔT2T1 MRI index and postoperative residual tumor computed on T2-weighted MR images.
All analyses were conducted using Stata/SE software (version 14.0 Stata Corp.), and data were presented as HRs and 95% CIs.
Informed consent. Considering that the study was retrospective, written consent to participate in the study was not applicable. Written informed consent was obtained for surgery.

Results
Study population. Three hundred and fifty-two adult patients underwent surgery at our institution for primary supra-tentorial LGGs. A total of 288 patients met the inclusion criteria. Patient demographic, clinical and radiological data are summarized in Table 1. According to clinical and neuropsychological status of each subject, AS was selected in 175 cases, GA was used in 113 cases.
Patients were surgically treated as follows: • Overall, new postoperative deficits were noted in a total of 100 patients (34.72%). The incidence of permanent postoperative deficits was relatively low in all protocol groups. Post-surgical deficits were stratified according to each surgical protocol ( Table 2).
In detail, new deficits with Protocol 1, 2 and 3 were noted respectively in 38.96%, 34.31% and 31.08% of cases in the immediate post-operative period and in 6.49%, 3.65% and 2.7% of patients 6 months after surgery.
No intra or postoperative mortality was observed.
EOR and surgical protocol. The median EOR was 90% (28-100) in the study population. The mean EOR resulted to be statistically different between patient operated with AS when compared to those operated under GA (p = 0.006). Among patients operated on AS, mean EOR was higher in those in which the Protocol 3 was applied (p = 0.001) (Fig. 1). Figure 1D stratifies the EOR subgroups according to the intraoperative protocol. A significant association was found between postoperative tumor volume on T2-weighted MRI images and ΔT2T1 value (rho = 0.788, p < 0.001).
Specifically, results of multivariate analysis showed that OS is independently associated with preoperative tumor growing pattern expressed by ΔT2T1 (p = 0.004), molecular class (p = 0.028), EOR (p = 0.018) ( Table 4, Fig. 2). The intraoperative protocol was not significantly associated with OS after multivariate analysis probably as a consequence of the more overwhelming effect of EOR. Figure 2 shows the OS Kaplan-Meier curves stratified according to preoperative tumor volume, ΔT2T1, EOR, volume of residual tumor, molecular class and intraoperative protocol.
Data on 3-, 5-, and 10 years estimated OS, PFS and MPFS were also stratified according to the EOR achieved and molecular class (Tables 5, 6).

Discussion
In this retrospective study, which included 288 adult patients surgically treated for DLGG, the use of different intraoperative protocols implemented in a large single center series over time was assessed based on neurological outcomes, EOR and OS.
The most relevant results we focused on were: (1) The evolution of the intraoperative surgical protocol over time led to an improvement in EOR.
(2) The highest percentage of total resection (EOR 100%) was obtained amongst patients treated with Protocol 3 (DTI and fMRI data loaded in NN + AS + RTNT). (3) Preoperative ΔT2T1 MRI index, EOR and the molecular class are confirmed as independent predictors for OS. (4) Tumor recurrence was also seen in patients that underwent radical resection and amongst all molecular classes.
The role of EOR. In the last decades, DLGG management paradigms have evolved from "wait and see" strategy to a more active interventional approach that aims at reducing the risk of malignant transformation. Numerous studies strongly support that a more extensive resection of DLGGs is associated with improved overall survival time and tumor progression free survival time 1, 3, 6, 11, 12, 24-38 .  Our results confirm that EOR is the strongest independent predictor for OS in DLGG: patients who underwent radical resection (EOR = 100%), when functionally possible, showed significant survival benefit.
Association between postoperative tumor volume on T2-weighted MRI images and preoperative ΔT2T1 MRI index showed that the greater the preoperative ΔT2T1 MRI index, the less extensive was the possible resection; it highlights a possible predictive correlation between ΔT2T1 MRI index and EOR achievable as previously demonstrated 38 .
A major volumetric difference between T2-weighted and contrast-enhanced T1-weighted MRI sequences suggests a greater propensity of the tumor to have a diffuse growing pattern and consequently to be less resectable.
The intraoperative surgical protocol. The principal limiting factor of surgical radicality lies in the infiltrating nature of these tumors, which gradually overgrow into eloquent areas both at cortical and subcortical level 14,40 . A personalized anatomo-functional planning is consequently mandatory 6,17 .
In the modern era of glioma surgery, increasing technological advances in imaging methods (such as intraoperative NN, MRI, and ultrasonography), stimulation mapping, and/or somatosensory-evoked potential techniques, are enabling surgeons to maximize cytoreduction and minimize morbidity.
Although there is growing literature supporting the accuracy and efficiency of individual techniques and methods 6,11,41 , we found no studies in literature that report the analysis of multimodal integrated protocols in the same cohort of DLGGs.  www.nature.com/scientificreports/ Volumetric studies have shown the prognostic significance of EOR for survival outcomes but have not compared the impact of the various intraoperative imaging modalities and surgical strategies in assisting/optimizing the maximal safe resection 13,18 .
This investigation assessed the efficacy of the combination of different techniques in surgery of DLGGs that are positioned in or close to eloquent areas both in terms of morbidity and EOR.
Specifically, neuroradiological techniques, such as fMRI and 3D reconstruction of white matter tracts, represent useful tools to preoperatively analyze the three-dimensional relationships between the tumor and the www.nature.com/scientificreports/ neighboring cortical functional areas and subcortical pathways, respectively 6,18,[42][43][44] . Intraoperative use of a guided navigation system enriched with functional and DTI data used in Series 2 significantly increase the possible EOR. Patients in Series 1 had a mean EOR of 83%, while those in Series 2 had a mean EOR of 88% (p < 0.001).
Overall, the use awake mapping and RTNT (Protocol 3) resulted in a higher EOR and lower permanent deficits. The data show that Protocol 3 was superior to Protocol 1 and 2 in improving the EOR. The median EOR and the number of radical resections is hence superior in patients belonging to Protocol 3 in comparison with those www.nature.com/scientificreports/ belonging to Protocol 1 and 2 (p < 0.001) (Fig. 1). Indeed, RTNT used in Protocol 3 proved to be an important step of the brain mapping technique. Throughout surgical resection, the patient performs multiple complex tasks involving the cognitive function(s) related to the brain area where the surgeon is working at that moment, thus guiding the surgical team. When a significant decrement of performance is noted, the surgeon can stop resection or move to another area, otherwise the surgical resection can be carried out without the uncertainty of "classical" negative mapping. As a matter of fact, RTNT can be defined as a continuous neuropsychological monitoring.
Thus, a multimodal integrated protocol with those techniques (brain mapping and multimodal intraoperative imaging) and strategies (AS and RTNT) can improve the chances of an extensive resection, which has repercussions on survival and recurrence. In light of that, Protocol 3 could represent a feasible support to the surgeon to safely maximize EOR in DLGGs surgery.
DES, both at cortical and subcortical levels, according to some experts of this field of surgery, represents the standard of care for DLGGs resection. It is used both for brain mapping and for monitoring neurologic performance, often in awake setting 13 . Table 5. Overall survival, progression free survival and malignant progression free survival at 3, 5 and 10 years after surgery according to extent of resection. www.nature.com/scientificreports/ Our study showed that patients who underwent AS had a better EOR in comparison to those operated under GA (p = 0.006) supporting the importance of AS for DLGGs in eloquent areas. This is the sole technique enabling a direct identification of neural networks and thus maximizing the resection while minimizing the risk of permanent deficits 39 . RTNT is based on a continuous neuropsychological testing administered during the surgical resection in order to fill the functional information gap between two consecutive DES, thus aiming at exploiting at its maximum the potentiality of AS 17 .
With regards to the neurological outcomes, our results showed that patients belonging to Protocol 3 resulted in fewer late neurological deficits than those that underwent Protocol 1, even if the difference was not significant, which was mainly due to the limited sample size of this rare disease.
De Witt Hamer et al. hypothesized that AS enables more extended resection and improved tumor control, resulting in the preservation of neurological functions that can be mapped at the cost of early transient neurological deficits 13 . In this investigation, in line with previous studies, the simultaneous decrease in transient and permanent deficit testifies the preservation of functional subcortical pathways, ensuring the postoperative plasticity mechanism for the neurological recovery 45,46 .
The molecular landscape. The 2016 WHO classification incorporated molecular findings into brain tumor diagnosis. Phenotypic and genotypic parameters in the diagnosis of DLGGs (mutations in the IDH1 and IDH2 genes mutations and 1p/19q co-deletion) have become of crucial importance 47 .
This investigation confirms the role of molecular class as an independent predictor of OS 35,48 . It also confirms the association between the molecular class and EOR achieved 22 .
The median EOR was larger in Oligodendroglioma than in Diffuse Astrocytoma IDH1/2 wild-type (92% vs. 83%; p = 0.002) and tended to also be larger in Diffuse Astrocytoma IDH1/2 mutated than in Diffuse Astrocytoma IDH1/2 wild-type (87% vs. 83%; p = 0.05). This difference could be related to a different tumor growing pattern.
Unfortunately, our data showed that DLGGs recurred in some of the patients even in cases of radical resection (Table 6). Interestingly, 3 and 5 years estimated OS in patients with an EOR of 100% was higher in patients belonging to molecular class of Diffuse Astrocytoma IDH1/2 mutated and Oligodendroglioma. Otherwise, considering the 10 years estimated OS, the percentage was similar amongst Diffuse Astrocytoma IDH1/2 mutated and Oligodendroglioma, and greater for Diffuse Astrocytoma IDH1/2 wild-type.
Future studies are thus required to better investigate the rich set of pertinent genomic and molecular markers, in addition to the radiological features in the era of radiogenomics, to better understand the biological behavior of DLGGs.
Limitations. The retrospective nature of the investigation and the different treatments performed at tumor progression are the principal limitations of the present study. Nevertheless, we attempted to homogenize our data set by studying a large sample of patients based on strict inclusion criteria.
Furthermore, considering the low incidence of DLGGs, well-designed retrospective studies, even if limiting in nature, can contribute to the existing body of knowledge.
T2w-FLAIR sequences were not available for both pre-and postoperative measures for some patients. We used the late postoperative scan to minimize overestimation of postoperative tumor volume due to edema or ischemia and to minimize potential differences between sequences 49,50 . Gliomas were classified according to the currently used WHO 2016 classification in this study, however, in light of the wide heterogeneity of DLGG, future investigations based on texture features from multi-parametric MRI and next generation sequences (NGS) analysis, may better clarify tumor recurrence even for those cases with a radical resection 47,51 .
In addition, the cornerstone of LGGs surgery is based on the concept of maximal safe resection 5,17,39,40 . Although intraoperative mapping of brain areas was shown to promote greater extent of resection and reduce functional deficits, this was shown only recently for some non-invasive techniques. We recognize that in this clinical setting the preoperative functional mapping by navigated transcraninical magnetic stimulation (nTMS) is an increasing valuable tool to stratify the individual risk of surgery-related deficits 52,53 . Our presurgical workup, at the moment, does not include this technique that has become an important tool in the modern surgical glioma management.
In closing, considering the higher interindividual variability of eloquent area arrangement and the wide potential displacement of functional sites, both at cortical level and subcortical pathways induced by the DLGG growing, a personal tailored multimodal integrated assessment of functional area should be developed to adopt in DLGG surgical field.

Conclusions
Technological tools are clinically useful to improve the extent of resection. Multimodal integrated approaches combining awake craniotomy, RTNT, intraoperative brain mapping and functional imaging analysis can enhance maximal safe and efficient tumor resection with possible subsequent OS benefits.
Combining volumetric studies and Next Generation Sequence investigations will provide a better understanding of DLGG evolution, paving the way for molecular neurosurgery.